A Knowledge-poor Pronoun Resolution System for Turkish
This provides the first fully specified knowledge-poor computational framework for pronoun resolution in Turkish, addressing a gap due to Turkish's different structural properties compared to other languages.
The paper tackles pronoun resolution in Turkish by developing a knowledge-poor system that uses empirically determined constraints and preferences, achieving performance significantly better than a baseline algorithm.
A pronoun resolution system which requires limited syntactic knowledge to identify the antecedents of personal and reflexive pronouns in Turkish is presented. As in its counterparts for languages like English, Spanish and French, the core of the system is the constraints and preferences determined empirically. In the evaluation phase, it performed considerably better than the baseline algorithm used for comparison. The system is significant for its being the first fully specified knowledge-poor computational framework for pronoun resolution in Turkish where Turkish possesses different structural properties from the languages for which knowledge-poor systems had been developed.